[Users] Visit & CarpetHDF5 mismatch?
Ian Hinder
ian.hinder at aei.mpg.de
Wed Apr 11 12:20:03 CDT 2012
On 11 Apr 2012, at 19:10, Kelly, Bernard J. (GSFC-660.0)[UNIVERSITY OF MARYLAND BALTIMORE COUNTY] wrote:
> Hi Ian.
>
>
> I'll look for a logfile, but there was nothing generated in my ~/.visit
> that I could see.
>
> Which reminds me: can frequent Visit/CarpetHDF5 users confirm for me where
> they get (a) Visit, and (b) CarpetHDF5? For me the answers are:
>
> (a) the "Mac OS X - Intel 64 bit" binary from here:
> https://wci.llnl.gov/codes/visit/executables.html
>
> (b) the cactuscode VizTools trunk via svn: svn co
> https://svn.cactuscode.org/VizTools/CarpetHDF5/trunk/ CarpetHDF5
>
> The instructions in the latter's README.txt seem a bit out of date, which
> made me suspicious. The lines about "updating the paths to the HDF5
> library", for instance:
Is the Carpet VisIt plugin part of the ET? If it is, then we should make sure the documentation (i.e. the README) is up-to-date for each ET release. If not, we should probably do it anyway :)
>
> install_name_tool -change '@executable_path/../lib/libhdf5.dylib'
> '/usr/local/lib/libhdf5.1.dylib'
> ~/.visit/darwin-i386/plugins/databases/libEvisitCarpetHDF5Database_ser.dyli
> b
>
> Now my 64-bit installation generates three subdirectories in ~/.visit:
> "2.4.2", "darwin-i386", and "darwin-x86_64". Of these, only "2.4.2" had
> any content in the "plugins/databases/" subdirectory, and these dylibs
> were *similar but not identical* in name to what's mentioned in the README
> (e.g., libECarpetHDF5Database_ser.dylib rather than
> libEvisitCarpetHDF5Database_ser.dylib).
>
> So ... should I be using a different version of Visit (older, 32-bit,
> etc.)?
>
> Bernard
>
> On 4/10/12 4:24 PM, "Ian Hinder" <ian.hinder at aei.mpg.de> wrote:
>
>>
>> On 10 Apr 2012, at 22:21, Kelly, Bernard J. (GSFC-660.0)[UNIVERSITY OF
>> MARYLAND BALTIMORE COUNTY] wrote:
>>
>>> Hi all. I'd like to visualise some 3D data produced by an initial-data
>>> run
>>> of the EinsteinToolkit (Maxwell).
>>>
>>>
>>> I'm outputting using CarpetIOHDF5; I have an output file
>>> "hahndol::hahndol_coulomb.h5", and decided to use Visit, as recommended
>>> by
>>> cactuscode, etc.
>>>
>>> I grabbed the CarpetHDF5 plugin code from svn, and installed as per
>>> instructions (pointing at the ETK-installed HDF5 in my
>>> Cactus/configs/<config-name>/scratch/external directory).
>>>
>>> Starting Visit (2.4.2) , and trying to load a .h5 file yields the
>>> following error:
>>>
>>>
>>> -------------------------------------------------------------------------
>>> --
>>> ----
>>> The metadata server running on host localhost has exited abnormally.
>>> VisIt
>>> is trying to restart it.
>>> Shortly thereafter, the following occured...
>>> VisIt was unable to open
>>>
>>> "/Users/bjkelly1/Projects/CODES/Cactus/exe/BL_IJS_test/hahndol::hahndol_c
>>> ou
>>> lomb.h5". Each attempt to open it caused VisIt's metadata server to
>>> crash. This can occur when the file is corrupted, or when the
>>> underlying
>>> file format has changed and VisIt's readers have not been updated yet,
>>> or
>>> when the reader VisIt is using for your file format is not robust.
>>> Please
>>> check whether the file is corrupted and, if not, contact a VisIt
>>> developer.
>>>
>>> -------------------------------------------------------------------------
>>> --
>>> ----
>>>
>>>
>>> I have no idea whether the file is corrupted, but wanted to know: is the
>>> CarpetHDF5 plugin up-to-date with the output of Maxwell's CarpetIOHDF5?
>>>
>>> Any basic ideas where I might have gone wrong?
>>
>> One problem can come from Visit trying to open too many files for the
>> maximum open file limit of your OS. Mac OS, for example, has a very
>> small limit. Try changing this with ulimit. But ultimately, you should
>> try to find the actual error message. I think there is an option for
>> Visit to output the errors to a file, but it's so long since I've used
>> Visit that I don't remember. Probably someone else can fill that in...
>>
>> --
>> Ian Hinder
>> http://numrel.aei.mpg.de/people/hinder
>>
>
--
Ian Hinder
http://numrel.aei.mpg.de/people/hinder
More information about the Users
mailing list